Slow movement small target detecting method base on frame correlation

摘要

For inter-frame motion characteristics of slowly moving small target, this paper is proposed a target detection algorithm based on relevant frames. This method use improved spatial high-pass filter for image pre-processing to improve signal to noise ratio of small targets; extract candidate target point by binary segmentation, and then conduct inter-frame correlation to binary image. Correlation is inter-frame corresponding to pixels in a neighborhood, and results stored in the tag matrix m, the matrix elements stand for the times of successful correlation. Here we use the rules of k/n rule to determine the output of the tag matrix, that is, if the n frames has k frames correlation successful, then the point is a candidate target. The advantage of this approch is: the short-term loss of target in the image sequence does not affect the detection effect, at the same time, overcome the high noise impact on small targets. Test results show that the method is a robust and efficient approach.